EMLIO: Minimizing I/O Latency and Energy Consumption for Large-Scale AI Training Conference: ACM/IEEE Conference; November 2025; St. Louis, MO, USA CCS: Computer systems organization Cloud computing CCS: Computer systems organization Energy-efficient computing CCS: Computing methodologies Machine le
Hasibul Jamil, MD S Q Zulkar Nine, Tevfik Kosar · 2025
Read EMLIO: Minimizing I/O Latency and Energy Consumption for Large-Scale AI Training Conference: ACM/IEEE Conference; November 2025; St. Louis, MO, USA CCS: Computer systems organization Cloud computing CCS: Computer systems organization Energy-efficient computing CCS: Computing methodologies Machine le by Hasibul Jamil, MD S Q Zulkar Nine, Tevfik Kosar (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-11035.txt.
Description
Hasibul Jamil, Zulkar Nine and Tevfik Kosar built in 2025 EMLIO, a service that feeds data to large AI training runs while cutting both waiting time and energy use. As datasets outgrow local disks, loading across networks becomes the bottleneck, and most loaders ignore its power cost. Their design places a light server on storage machines that batches samples and streams them ahead of need. A graphics-side library then prepares the data for the model.
Across local, office and wide-area links, the service loaded data up to 8.6 times faster than leading loaders. Energy use fell by up to about eleven times. Performance stayed flat even as network distance grew, a rare trait. The authors offer the design as a blueprint for greener AI clouds. They argue energy belongs beside speed in judging data pipelines.
Catalog record
| CREATOR | Hasibul Jamil, MD S Q Zulkar Nine, Tevfik Kosar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-11035